Interpretation of GRACE data using a groundwater recharge model€¦ · Interpretation of GRACE data using a groundwater recharge model H. C. Bonsor et al. Title Page Abstract Introduction

HESSD7, 4501–4533, 2010

Interpretation ofGRACE data using a

groundwaterrecharge model

H. C. Bonsor et al.

Title Page

Abstract Introduction

Conclusions References

Tables Figures

J I

J I

Back Close

Full Screen / Esc

Printer-friendly Version

Interactive Discussion

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Hydrol. Earth Syst. Sci. Discuss., 7, 4501–4533, 2010www.hydrol-earth-syst-sci-discuss.net/7/4501/2010/doi:10.5194/hessd-7-4501-2010© Author(s) 2010. CC Attribution 3.0 License.

Hydrology andEarth System

SciencesDiscussions

This discussion paper is/has been under review for the journal Hydrology and EarthSystem Sciences (HESS). Please refer to the corresponding final paper in HESSif available.

Interpretation of GRACE data of the NileBasin using a groundwater rechargemodelH. C. Bonsor1, M. M. Mansour2, A. M. MacDonald1, A. G. Hughes2, R. G. Hipkin3,and T. Bedada3

1British Geological Survey, Murchison House, West Mains Road, Edinburgh, UK2British Geological Survey, Kingsley Dunham Centre, Keyworth, Nottingham, UK3Grant Institute, Edinburgh University, West Mains Road, Edinburgh, UK

Received: 3 June 2010 – Accepted: 5 July 2010 – Published: 9 July 2010

Correspondence to: H. C. Bonsor ([email protected])

Published by Copernicus Publications on behalf of the European Geosciences Union.

4501

http://www.hydrol-earth-syst-sci-discuss.net

http://www.hydrol-earth-syst-sci-discuss.net/7/4501/2010/hessd-7-4501-2010-print.pdf

http://www.hydrol-earth-syst-sci-discuss.net/7/4501/2010/hessd-7-4501-2010-discussion.html

http://creativecommons.org/licenses/by/3.0/

HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Abstract

Assessing and quantifying natural water storage is becoming increasingly important asnations develop strategies for economic growth and adaptations measures for climatechange. The Gravity Recovery and Climate Experiment (GRACE) data provide a newopportunity to gain a direct and independent measure of water mass variations on a5

regional scale. Hydrological models are required to interpret these mass variationsand partition them between different parts of the hydrological cycle, but groundwaterstorage has generally been poorly constrained by such models. This study focusedon the Nile basin, and used a groundwater recharge model ZOODRM (Zoomable Ob-ject Oriented Distributed Recharge Model) to help interpret the seasonal variation in10

terrestrial water storage indicated by GRACE. The recharge model was constructedusing almost entirely remotely sensed input data and calibrated to observed hydrolog-ical data from the Nile. GRACE data for the Nile Basin indicates an annual terrestrialwater storage of approximately 200 km3: water input is from rainfall, and much of thiswater is evaporated within the basin since average annual outflow of the Nile is less15

than 30 km3. Total annual recharge simulated by ZOODRM is 400 km3/yr; 0–50 mm/yrwithin the semi arid lower catchments, and a mean of 250 mm/yr in the sub-tropicalupper catchments. These results are comparable to the few site specific studies ofrecharge in the basin. Accounting for year-round discharge of groundwater, the sea-sonal groundwater storage is 100–150 km3/yr and seasonal change in soil moisture,20

30 km3/yr. Together, they account for between 50 and 90% of the annual water storagein the catchment. The annual water mass variation (200 km3/yr) is an order of mag-nitude smaller than the rainfall input into the catchment (2000 km3/yr), which could beconsistent with a high degree of moisture recycling within the basin. Future work isrequired to advance the calibration of the ZOODRM model, particularly improving the25

timing of runoff routing.

4502





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

1 Introduction

Water storage is a driver for economic growth, poverty reduction and a key compo-nent of a nation’s resilience to climate change (Grey and Sadoff, 2007; DFID, 2010).For many people in the world, accessing stored water means using groundwater ratherthan building dams and reservoirs (MacDonald and Calow, 2009). Therefore, as rainfall5

and surface water resources become less reliable with predicted climate change, beingable to assess the spatial distribution and seasonal variation of stored water masses –and in particular impacts on groundwater recharge – is increasingly important (Bates etal., 2007; MacDonald et al., 2009). Following the early demonstration that the GravityRecovery and Climate Experiment (GRACE) satellites were able to detect hydrologi-10

cally induced gravity changes (Tapley et al., 2004), GRACE data has become a widelyused tool for gaining an insight to regional scale water mass variations (e.g. Andersonand Hinderer, 2005; Brown and Tregoning, 2010). Processed GRACE data, correctedfor atmospheric and oceanic water variations, give an independent estimate of the totalmonthly change in terrestrial water storage within a region. The GRACE data is an15

estimate of bulk water storage changes, and the data cannot distinguish between thedifferent components of stored water mass (Rodell et al., 2004; Schmidt et al., 2006).GRACE data must, therefore, be interpreted using hydrological modelling. Despite thislimitation, GRACE data remains a highly valuable dataset as an independent measureof large-scale, bulk water storage variations.20

GRACE data have been used successfully to assist basinal-scale water balancecalculations (e.g. Rodell et al., 2004; Syed et al., 2005), and to quantify water massstorage changes (e.g. Rodell et al., 2009). To date, most published work interpretingGRACE data has been in regions where there are large mass storage changes related,predominantly, to a single stored water mass (e.g. the major river basins (Schmidt et25

al., 2006); regions of groundwater depletion in India (Rodell et al., 2009)). Bulk waterbalance models (e.g. the World Global Hydrological Model WGHM (Doll et al., 2003))have been widely used to interpret the mass storage changes indicated in GRACE

4503





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

quantitatively, with respect to the regional hydrological cycle. These bulk water balancemodels use climatic and land-use data to calculate the volume of rainfall, evaporatedwater and surface runoff within any given catchment. However, the models are unableto accurately model each component of terrestrial water storage. Soil moisture andgroundwater are often simply calculated as a “remainder” in the water balance models,5

leading to the conclusion that a major limitation of the models is their inability to modelchanges in groundwater storage (Schmidt et al., 2006). As a result, most work to datehas not harnessed the full potential of GRACE data to interpret changes in groundwaterstorage on a basinal scale. Being able to use GRACE data with hydrological modelsthat can simulate all components of terrestrial water storage, particularly groundwater10

storage, would therefore be of significant value.To interpret how seasonal water storage variations indicated in GRACE data relate to

changes in each component of water storage in a basin, requires the use of hydrologi-cal models that are capable of simulating the processes which lead to the partitioningof precipitation between evaporative losses, surface run-off, soil moisture and ground-15

water recharge on a basin-scale. Rainfall-runoff models (e.g. Beven, 2002) and bulkwater balance models concentrate on surface processes with a limited emphasis onsoil processes. For these models, recharge is only a “by-product” of moving wateraround the system, whereas recharge models are specifically designed to simulatesoil processes, especially the calculation of actual evaporation. Therefore, they are20

more appropriate tools to accurately simulate the variation of each component of ter-restrial water storage. Using recharge models it should therefore be possible to relateseasonal changes in different stored water masses to the bulk seasonal water storagechange indicated by GRACE data.

This study develops a recharge model for the Nile Basin, using the ZOODRM code25

(Hughes et al., 2008), and almost entirely remotely sensed input data. This rechargemodel is calibrated to observed hydrological data from the Nile, and then used to inter-pret the seasonal variation in terrestrial water storage indicated by GRACE data fromthe Nile Basin. Of prime interest in this study is the quantification of the proportion of

4504





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

the seasonal GRACE gravity mass variation in the Nile Basin that relates to groundwa-ter recharge and soil moisture changes.

2 The Nile Basin

2.1 The Nile Catchment

Approximately 3 million km2 in extent, the Nile River Basin drains almost 10% of Africa5

and includes over 10 countries (Sutcliffe and Park, 1999; Nicol, 2003). The distribu-tion of water resources in the Nile Basin is highly contentious, and water allocationsbetween neighbouring countries in the Nile Basin have historically been the cause ofmuch dispute (Nicol, 2003). The allocation of the Nile flow between the 10 countries inthe basin has, traditionally, been done solely on the basis of the proportion of annual10

surface water availability in each country. Controversy has arisen from this division ofthe Nile flows as a result of the large difference in climate and seasonal water availabil-ity between the upper and lower Nile catchment, and also as a result of the importanceof the water to livelihoods and industry within each of the countries (Conway, 2005).

Within Egypt, and much of Sudan, in the arid and semi-arid part of the catchment,15

the Nile River is the main source of water (Nicol, 2003). Division of the Nile River flowsis therefore incredibly important to these countries, not only for drinking and domesticwater needs, but for industry and economic growth (Karyabwite, 2000; Conway, 2005).Within the sub-tropical upstream catchment of the Nile Basin, seasonal rainfall some-what reduces water pressures in Southern Sudan, northeast Congo and south-west20

Ethiopia (Sutcliffe and Park, 1999). However, water pressures can still be high in thesecountries within the upstream catchment during the dry season.

More recently, there is growing interest in the total volume of water in each country,rather than the single focus on Nile flows (Karyabwite, 2000; DFID, 2010). Being ableto determine terrestrial water stores – such as groundwater – is therefore becoming25

increasingly important to water resource management in the Nile Basin. In the absence

4505





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

of hydrological data, or difficulty in obtaining such data, the ability to use remotelysensed data to develop accurate, independent hydrological models that can quantifyseasonal distributions of stored water mass in a basin, is potentially very important forwater resource management in the Nile (Brunner et al., 2007; Mileham et al., 2008;Abdalla, 2009). It is, therefore, in large transboundary basins, such as the Nile, that5

GRACE data can have highest impact potential as an independent measure of bulkterrestrial water storage variations.

2.2 Hydrology of the Nile Basin

The hydrology of the Nile Basin is complex. The hydrology of any particular reach isdirectly influenced by the adjacent landuse (e.g. irrigation abstraction) and the hydro-10

logical regime of major tributaries (Sutcliffe and Park, 1999).The Nile River extends from the head waters of Lake Victoria and Lake Albert in

Uganda and the Democratic Republic of the Congo, to the Mediterranean Sea – Fig. 1.Annual outflow to the Mediterranean is approximately 28 km3/yr. The basin is dividedinto a number of sub-catchments: the Victoria Nile, Sudd, White Nile, Sobat, Blue Nile15

and the Main Nile – Fig. 1. The contribution of each of these sub-catchments to theMain Nile flow is markedly different. Whilst the Victoria Nile and White Nile, providea constant annual baseflow to the Main Nile of approximately 60 km3/yr, as much as70% of the annual discharge of the Main Nile is sourced seasonally from the Blue Nile,Sobat and Atbara tributaries which originate in the Ethiopian Highlands (Sutcliffe and20

Park, 1999). These tributaries respond relatively rapidly to the wet season rainfall inthe Ethiopian Highlands and provide a seasonal pulse of water to the Main Nile Riverin the lower catchment. This seasonal response is not observed within the White Nile,because as much as half of the inflow from the Victoria Nile and Bahr el Ghazel sub-catchments is lost to evaporation within the Sudd wetland (Sutcliffe and Park, 1999) –25

Fig. 1.Evaporative losses of water throughout the Nile catchment, from both open water

evaporation and evapotranspiration, are known to be highly significant to the hydrology4506





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

of the Nile, and are evident by the reduction in flow down the Nile (Sutcliffe and Park,1999). In Sudan and Egypt, in the arid part of the catchment, open water evaporativelosses are 1700–2400 mm/yr, and the discharge of the Nile River actually decreasesdownstream as a result of evaporation (Mohamed et al., 2004; Sutcliffe, 2005; Vallet-Coulomb et al., 2001; Sutcliffe and Park, 1999) – Fig. 1.5

3 Nile GRACE data

The satellite mission Gravity Recovery and Climate Experiment (GRACE) has pro-vided independent monthly models of the Earth’s gravity field since 2002 (Tapley et al.,2004). GRACE lacks spatial resolution better than a few hundred kilometres (sphericalharmonic degree ∼ 50) but, when averaged over regions of this size, it detects con-10

sistent mass changes equivalent to the effect of a few millimetres of water (Bedada,2007). Gravity changes due to short period geological events, like volcanic eruptionsor earthquakes, are detectable in extreme cases but produce a step-like change pre-ceded and followed by a slow build-up or recovery (Han et al., 2006). Processes in thehydrosphere, atmosphere and cryosphere are believed to be the only ones capable of15

generating the mass changes with the seasonal timescale observed in the variabilityof the gravity field. Month to month changes seen in the gravity field over a low latitudecontinental region correspond to changes in all sources of sub-satellite water, inte-grated over the whole vertical column including aquifers, flooding and surface water,rainfall and other moisture in the atmosphere.20

For this study of the hydrology of the Nile Basin, GRACE-2 gravity data have beensynthesised from the CNES model (Lemoine et al., 2007; Bruinsma et al., 2010), al-though we have made comparisons with other models generated by groups in theTexas Centre for Space Research, the Geoforschungszentrum in Potsdam and the JetPropulsion Laboratory in Pasadena. The CNES model combines the GRACE satellite25

orbit data with a background gravity model (EIGEN-GRACE-02S) and accelerometerand attitude data, to generate stable solutions of gravity data which are corrected for

4507





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

the background solid earth mass variations (Lemoine et al., 2007). The second re-lease of GRACE data from CNES – GRACE-2 – used in this study, have also beencorrected for water mass changes occurring within the atmosphere and ocean, usingocean and atmospheric mass variation models (Lemoine et al., 2007). These modelssimulate oceanic and atmospheric water mass variations, respectively, according to5

climatic data and tides, and the estimated background variations in water mass varia-tion are subtracted from the total water mass variation indicated by GRACE data in aregion. As a result, gravity mass changes indicated by GRACE-2 data, should relateonly to monthly changes in terrestrial water storage in the region of interest, plus anyerror within the GRACE data incurred from the corrections. Based on comparison to10

gauged meteorological data the error within the Nile Basin, the atmospheric correc-tion is thought to be much less than 20 mm water equivalent in this region (Bettapdur,personal communication). Note that to predict monthly water mass changes from theGRACE gravity models, time changes of gravity must be identified within GRACE datafrom an arbitrary datum. This datum is defined by the particular month chosen as ref-15

erence. In this study March 2004, the driest hydrological month of the modelled period,was used as the reference month.

The Nile GRACE data from 2003 to 2005 indicate a regular seasonal variation ofstored water mass in the Nile Basin, equivalent to ±300 mm water – Fig. 2. Areas ofmass addition correspond with the location and timing of the highest seasonal rainfall.20

Across the Nile Basin, the seasonal change in water mass is approximately 150 km3

(estimated for 2004, Fig. 3).

4 Comparison of GRACE data with observed hydrological data

Figure 3 shows a comparison between monthly GRACE data (interpreted as thechange in the terrestrial water mass) and observed hydrological data of monthly rain-25

fall and river outflow from the Nile delta. The comparison of GRACE data to observedmonthly rainfall confirms that a significant volume of rainfall becomes stored within the

4508





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

basin. Annual mass change is approximately 150 km3, and total annual rainfall approx-imately 2000 km3. The small outflow of River Nile, relative to the monthly change inwater storage indicated by GRACE, clearly indicates only a very small proportion ofthe rainfall water becomes stored within the Nile River itself. It is inferred, therefore,that much of the monthly change in water storage indicated by the GRACE data re-5

flects seasonal storage of water in soil moisture, groundwater recharge and, surfacelakes and reservoirs. As GRACE data give only a bulk estimate of the change in waterstorage in the basin, hydrological modelling is required to gain a better understandingof which of these terrestrial water mass stores receive the most rainfall, and how eachcomponent of water storage changes seasonally within the basin.10

The observed hydrological data also indicate that the annual discharge of the NileDelta of 28 km3 is less than 2% of the annual rainfall (approximately 2000 km3) inputto the basin. This suggests a large volume of the rainfall reaching the ground surfacein the Nile Basin is lost through evapotranspiration. This evaporated water could beretained within the atmosphere, lost, or re-precipitated and recycled within the basins15

hydrological cycle. The magnitude of the correction of GRACE data for atmosphericwater in the latest release of data (the data used in this paper), gives an indicationas to which of these possibilities is most likely in the Nile Basin. Corrected GRACEdata indicate the monthly change in terrestrial water storage in a region; uncorrectedGRACE data (earlier releases of data) indicate the monthly change in total water mass20

(including atmospheric water). In the Nile Basin, there is, on average, only a 5–10%difference between the corrected and uncorrected GRACE data. This suggests thatthere is not a significant input of atmospheric water to the Nile Basin. It appears,therefore, that most of the atmospheric water within the Nile Basin is derived from localevaporation and evapotranspiration. The small difference between the corrected and25

uncorrected GRACE data indicates the monthly change in atmospheric water is notsignificant compared to other water stores, suggesting that much of the evaporatedwater must re-precipitate within the basin, rather than moving in or out of the basin.

4509





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

The ECMWF correction of GRACE data is believed to be accurate to well within20 mm water equivalent in the Nile Basin on the basis of the comparison betweenthe modelled atmospheric data to which the GRACE data is corrected, and observedmeteorological data (Bettadpur, personal communication). The inference that mostof the atmospheric water within the Nile Basin is derived from local evaporation is5

supported by work by Fontaine et al. (2002) in West Africa basins. Comparing detailedatmospheric modelling to gauged meteorological data, Fontaine et al. (2002), inferthat much of the atmospheric water in the West African basins is derived from localevaporation.

Comparing the Nile GRACE data to observed hydrological data is a useful exercise10

and clearly indicates that: (1) up to 10% of rainfall in the catchment becomes storedwithin the basin; (2) little water is stored within the River Nile, and therefore, a significantamount of rainfall must become stored within other terrestrial water stores – namely,soil moisture, groundwater recharge, and surface water bodies; (3) evaporative lossesare large within the Nile Basin, and a significant amount of the atmospheric water15

within the Nile Basin is sourced from local evaporation; and (4) much of the evaporatedwater is re-precipitated within the basin. GRACE data does not distinguish where watermass is stored in a basin. To be able to interpret which component of terrestrial waterstorage is most significant in the Nile Basin, therefore, requires the use of groundwaterrecharge modelling.20

5 ZOODRM groundwater recharge model

5.1 The ZOODRM model

The ZOODRM model is a distributed recharge model code for calculating spatial andtemporal variations in groundwater recharge, and has been applied successfully insemi-arid areas (e.g. Palestine-Hughes et al., 2008 and Inner Mongolia – O Dochar-25

taigh et al., 2010) and in wet temperate areas (e.g. Europe – Jackson et al., 2005).

4510





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

ZOODRM calculates recharge by applying the soil moisture deficit (SMD) rechargemethod over daily time steps (Penman, 1948; Grindley, 1967). The SMD method de-termines how the actual evaporation (AE) is related to the potential evaporation (PE) bythe amount of water stored in the soil at any time. Using a daily time step, rainfall lessrunoff, is added to the soil store and the appropriate evaporation is taken off. When the5

SMD is between zero and a value know as the root constant (C) then AE is equal toPE. However, once the SMD is greater than the root constant, but less than the wiltingpoint (D) then AE is a proportion (usually 10%) of PE. Once the SMD is above thewilting point then no further evaporation occurs as the plant is deemed stressed andcan no longer transpire. This method allows the impact of water available in the soil on10

the effects of plant water uptake to be assessed. Separate objects are used to repre-sent different entities within the ZOODRM code such as soil, rivers and springs. Thesenodes are, in turn, held in a two layer grid structure comprised of an unsaturated andsaturated grid. The ZOODRM model is therefore able to represent the numerous flowprocesses controlling the partitioning of surface water and groundwater, as well as the15

delay time required for water in the unsaturated zone to reach the water table (Hugheset al., 2008). ZOODRM is also able to simulate evaporative losses from open watersurfaces, such as the Sudd wetland, which are known to be important to the hydrologyof the Nile Basin (Sutcliffe and Park, 1999). ZOODRM outputs monthly estimates ofrainfall, evapotranspiration (ET), change in soil moisture, groundwater recharge, sur-20

face runoff and river discharge.Lerner et al. (1990) suggest that the SMD method can be used where the input to the

system, i.e. precipitation and irrigation returns, is greater than 500 mm/a. This conditionis reached, on average, for the Nile basin. However due to the spatial distribution ofrainfall, the condition is not met in the arid northern subcatchments. Here, rainfall and25

recharge is negligible and accounts for little of the overall Nile Basin mass balance.Therefore, a pragmatic approach was taken to apply the SMD over the whole basinwith the provision that the model does not produce anomalous results in the arid north(and therefore affect the mass balance of the basin).

4511





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

The ZOODRM model is a suitable model for this study due to its lower data demandsrelative to other hydrological models, the ability of the model to use largely remotelysensed data, and, the added functionality of routing of runoff water according to topog-raphy. The data needs of the model are: daily rainfall and potential evaporation (PE),land use, topographical aspect directions, geological and digital elevation data. Due to5

the size of the Nile Basin and difficulties in obtaining recorded rainfall and evaporationdata from individual countries, largely remotely sensed input data were used for themodelling work. Daily rainfall (NOAA data) and PE were sourced from the FEWS NETAfrican Dissemination Service for 2003–2005. Geological data were sourced from theDigital Geological Map of the World (UNESCO, 1991; Persits, 1997), land-use data10

from the USGS (http://edc2.usgs.gov/glcc/glcc.php; Loveland et al., 2000) and eleva-tion data from the Shuttle Mission (STRM) (http://www2.jpl.nasa.gov/srtm/; Farr et al.,2007).

A 20×20 km cell size was adopted for the model grid as a compromise betweenmodel accuracy and model run time. This model resolution, although coarse, is greater15

than that achieved by previous basin-scale modelling work in the Nile (e.g. Mohamedet al., 2005, whose modelling work was of 50×50 km resolution), and by other hydro-logical models used with GRACE data (e.g. Fukuda et al., 2009). All input data weregridded to be of the same 20 km2 resolution and projected to a UTM zone 36 projec-tion to ensure no error was introduced into the ZOODRM model when input data files20

were read. This projection of input data was required to ensure minimal scalar, orareal distortion of the data within the modelled area. Preservation of area within themodel is important to model accuracy, as the ZOODRM model integrates water fluxesover area. Due to the size of the Nile Basin, the curvature of the Earth would inducesignificant error to the area of the basin presented within conformal or equidistant pro-25

jections. Surface routing of runoff water is calculated according to a slope aspect mapgenerated from digital elevation model data (DEM). The aspect direction at any gridpoint determines the direction of movement of surface runoff, which is generated as aproportion of the effective precipitation plus any water received from an adjacent cell.

4512





http://edc2.usgs.gov/glcc/glcc.php

http://www2.jpl.nasa.gov/srtm/

HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

The recharge model has a simplified method of calculating run-off routing based on atime step of one day (Hughes et al., 2008). Currently this is fixed and is adequate forsmall sub-catchments. It is recognised that the timing of run-off would be greatly im-proved by allowing a variable timing and this would be a further refinement of the Nilebasin model. Calculation of recharge according to the soil moisture deficit method,5

relies on the root constant (C) and wilting point (D) values which are characteristicsof the vegetation type. For the different vegetation types within the Nile Basin, C andD values were obtained from published values (see FAO; Lerner et al., 1990). Othermodelling factors were kept as simple as possible in this preliminary stage of model de-velopment. Irrigation losses were set to zero to ensure that it was only the partitioning10

of surface water to run-off and recharge which determined the modelled river dischargeand groundwater water mass. Open water evaporation rates from the Sudd wetland,and the Nile river surface were modelled as 3–7 mm/day according to published values(e.g. Sutcliffe, 2005; Mohamed et al., 2004).

5.2 Model calibration15

The ZOODRM model was run for three full hydrological years from 2003 to 2005, andcalibrated to observed annual discharge from 1976–1979 (no modern data available) at10 gauging stations along the Main Nile and tributaries. No modern discharge data isavailable from the Nile Basin beyond 1979, bar two gauging stations on the White Nileand Main Nile. The discharge data from these stations, is very similar to the discharge20

data from 1976–1979 (±14% difference), indicating the use of discharge data from1976–1979 is valid in the absence of any other data.

The recharge model simulates observed annual discharge of the Nile River to within40% (and in most cases within 20%), using realistic values of run-off (modelled tovary seasonally between 10–27% of rainfall), overland losses (2% of run-off) and river25

losses (0–3% river discharge) – Table 1. These modelled values of run-off are compa-rable to other published estimates (e.g. Mohamed et al. (2005), who used runoff valuesof 5–29% rainfall). It was deemed valid to include overland losses (simulated by the

4513





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

RunOn parameter within the recharge model) due to the semi-arid nature of the lowercatchment and overland losses have proved to be important to previous ZOODRMmodels within semi-arid areas (e.g. Hughes et al., 2008). The overland losses (mod-elled as 2% of run-off) were calibrated according to comparison of the modelled dis-charge against the observed data available.5

Greatest error between observed and modelled river discharge occurs in the simu-lation of river flow within the Sudd wetland at Bahr el Jebel, in the upper catchment(modelled river flow is 43% greater than observed). Elsewhere, the model simulatesobserved annual discharge to within 20%. Based on this comparison of modelled andobserved annualised discharge in the Nile Basin, it is inferred that the ZOODRM model10

simulates the discharge of the Nile River and consequently the wider hydrological func-tioning of the Nile Basin to an acceptable level of accuracy. The water balance of themodel, in which no long-term, inter-annual, gain or loss of water is modelled in thebasin, also indicates the model is replicating the observed hydrological functioning ofthe Nile Basin accurately. It is known that there are no large-scale abstractions of wa-15

ter in the Nile Basin (irrigation abstractions being of relatively low intensity comparedwith other countries worldwide), or major changes in rainfall (Sutcliffe and Park, 1999).GRACE data also do not indicate a long-term interannual change to the total watermass within the basin between 2003 and 2005. However, undoubtedly, a better cal-ibration of the ZOODRM model would be facilitated by access to modern discharge20

data.

5.3 Modelling results

The ZOODRM model simulates the partitioning of the observed rainfall mass betweenother components of the water cycle in the river basin, and calculates the daily quan-tities of evapotranspiration (ET), runoff, river outflow and groundwater recharge in the25

Nile Basin throughout each hydrological year from 2003–2005. Simulated monthly vol-umetric estimates of ET, runoff, river outflow and groundwater recharge in the Nile aredisplayed in Fig. 4. The modelling results replicate both observed data, and the known

4514





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

hydrological functioning of the Nile Basin within acceptable limit of error (i.e. plausiblerunoff, recharge and evaporation are modelled). This indicates that at a basic levelthe recharge model simulates the key hydrological processes in the Nile Basin suffi-ciently well so that the modelled functioning of the Nile Basin matches the observeddata available.5

Modelling results confirm that river outflow from the basin is only 1.5–2% of the ob-served rainfall, and consequently, the Nile River accounts for only a very small amountof the water stored within the Nile Basin. This inference is supported by the comparisonof GRACE data to observed hydrological data. The volume of water lost from the NileBasin by river discharge is also small relative to the volume of water evaporated within10

the basin; the annual outflow of the Nile River being less than 15% of the evaporatedwater mass.

Modelling results indicate evaporative losses to be very large within the Nile Basin.The large difference between the initial volume of surface runoff generated by rainfall,and the annual discharge of the Nile River, suggests a large proportion of runoff water15

becomes evaporated (∼8 km3/yr).A significant volume of rainfall is modelled to become stored as groundwater.

Groundwater recharge is modelled to be greatest within the Ethiopian and SudaneseHighlands and northeast Congo, corresponding to the areas of highest rainfall in theNile Basin – Fig. 5. High recharge is also observed within 20 km of the Nile River20

due to direct recharge through the river bed – river losses were modelled to be 0–3% depending on geology and based on river losses estimated from isotope studies(e.g. Abdalla, 2009). Within the semi-arid lower catchment average annual rechargeis modelled to be 0–50 mm/yr, whilst in the sub-tropical upper catchment modelledrecharge is on average 250 mm/yr. The total annual groundwater recharge simulated25

by ZOODRM is 400 km3/yr, which is considerably greater than river outflow from theNile Basin. This estimate is comparable to other larger scale recharge estimates inAfrica by Doll and Fiedler (2008) who calculate 0–200 mm/yr of recharge in the NileBasin on a 0.5 degree grid, equivalent to 50×50 km resolution. The recharge estimate

4515





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

is also comparable to those from smaller regional studies in the Nile which use iso-tope and chloride concentrations as indicators of groundwater recharge and discharge– Fig. 5. Taylor and Howard (1996) determined annual groundwater recharge in theAroca catchment of the Victoria Nile, in central Uganda, to be ∼200 mm/yr, using a soilmoisture balance model and isotope data. The modelled recharge value for the same5

region by this work is ∼150–200 mm/yr. MacDonald et al. (2009) determined ground-water recharge in the Singida region of northern Tanzania to be 10–50 mm/yr, which iswithin 20% of the recharge modelled by this work. Within southern Sudan, Abdalla etal. (2009) determined recharge from direct infiltration of rainfall through the soil to beless than 10 mm/yr at distances 20–30 km away from the Nile River. Recharge values10

modelled for the same region by the ZOODRM model are 0–50 mm/yr. Finally, withinnortheast Sudan, Farah et al. (1999) determined the contribution of modern rainfall togroundwater recharge, near the capital city, Khartoum, to be minimal. This inference issupported by the results of this modelling work, which estimates 0–50 mm/yr of modernrecharge at Khartoum, with values of more than 40 mm annual recharge being atypical.15

6 Discussion

6.1 Moisture recycling in the Nile Basin

This section discusses the relationship of GRACE data to moisture recycling resultingfrom the cycling of rainfall and evaporation in the same catchment. GRACE data in-dicate an average seasonal water mass change of approximately 200 km3 within the20

Nile Basin each year; annual rainfall is approximately an order of magnitude greater,2000 km3. Therefore, 90% of rainfall is not retained, but is lost either through the Nileinto the Mediterranean Sea, or to the atmosphere through evapotranspiration. Sinceaverage annual Nile flow at the Mediterranean is only 28 km3, virtually all the terrestrialwater lost to the atmosphere through evapotranspiration.25

4516





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

However, annual changes in atmospheric water volumes are inferred to be less than10% of the total annual water change by the comparison of corrected and uncorrectedGRACE data using the ECMWF model. This suggests that much of this evaporatedwater of is rapidly re-precipitated within the basin, rather than the water being held inthe atmosphere. This is in contrast to the work of Mohamed et al. (2005) who esti-5

mated much lower (11%) moisture recycling within the Nile Basin. However, a similarlarge influence of moisture recycling within West African river basins was suggestedby Fontaine et al. (2002), based on work with atmospheric models and observed me-teorological and hydrological data.

6.2 Groundwater and soil storage10

Groundwater recharge modelled by ZOODRM is a calculation of the total volume ofwater draining from the soil into the ground. However, assuming a quasi-equilibriumannual water balance in Nile Basin (i.e. no large long-term water deficit or surplus) wecan infer there is neither a large net gain or loss of groundwater. This assumption isreasonable for the basin since regional groundwater abstraction is limited. Under this15

assumption, annual recharge will be approximately matched by an equal volume ofgroundwater discharge within the Nile Basin. Discharge will occur largely as baseflowto rivers, groundwater abstraction and direct evapotranspiration from shallow ground-water (Farah et al., 1999) and it is likely, therefore, to be relatively constant throughoutthe year. We have used the term ∆ groundwater storage to refer to the difference20

between modelled monthly groundwater recharge and monthly groundwater discharge(which is assumed constant each month and therefore equal to the average monthlygroundwater recharge). Using this approach, the seasonal mass change in the NileBasin due to groundwater is estimated to be 100–150 km3/yr for 2003–2005 (Fig. 6).The seasonal change in groundwater storage accounts for more than 50% of the an-25

nual water mass variation calculated from GRACE data.Modelled changes in soil moisture throughout the year are approximately 30 km3

and are most significant at the beginning of the rainy season as the soils wet up before4517





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

significant groundwater recharge can occur, and at the beginning of the dry season asthe soils dry out.

Taken together, the simulated annual variation in groundwater storage and soil mois-ture accounts for 50–90% of the variation in stored water mass indicated by GRACEdata in the Nile Basin – Fig. 7; the remainder of water mass is runoff. As a result5

of the slow river flows, irrigation, and high evapotranspiration rates in the Nile basin,much of the runoff is evaporated within several months rather than flowing out into theMediterranean. In wetter years (e.g. 2003) there is a much higher runoff, and a greaterproportion of storage is not accounted for by groundwater and soil storage – Fig. 7.

6.3 Limitations of modelling results10

There are three main limitations to the ZOODRM recharge model for the purposesof this work: (1) ZOODRM simulates run-off to be lost too quickly to the River Nile,and therefore the volume of run-off is generally underestimated by the model; (2) theZOODRM model underestimates the storage of surface water (lakes, wetlands) andsimulates the seasonal variation in this surface water store poorly; and (3) storage of15

water in artificial reservoirs is excluded from the model.

Simulation of runoff

As described in Sect. 5.1, ZOODRM uses a fixed time step of one day and routes waterdown the topographic gradient during the time step. Therefore, there is no delay in rain-fall striking the ground surface and reaching the river channel. This means that run-off20

arrives at and discharges from the river system too quickly. This has implications for thecomparison of the water stored in the system that is calculated by the recharge modelwith the GRACE data. Within particularly wet years (e.g. 2003), when the volume ofrun-off to the Nile catchments water balance will be more significant, the underestima-tion of run-off and the resulting surface water storage, is most apparent when the bulk25

4518





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

stored water mass modelled in the catchment is compared to GRACE data – Fig. 7.Within drier years (e.g. 2004) when there is less rainfall, and run-off is less significant,the bulk stored water masses modelled by the recharge model compare much betterto the annual water storage variation indicated by the GRACE data. This limitation ofthe model is not, however, thought to incur significant error to the calculation of soil5

moisture and recharge volumes, as the model calculates theses water masses usingthe soil moisture deficit approach, and the most significant factors to the calculations –the volume of effective precipitation reaching the ground surface, potential evaporation(PE) and land cover – do not rely on the run-off.

As a result of the ZOODRM recharge model routing run-off water into the Nile River10

too quickly, the model underestimates the volume of run-off water which will be evap-orated, to be either lost from the basin, or more likely, to be re-precipitated. Thismodelling limitation is unlikely to be significant to the recharge calculated by the model,but it is important to note as a model limitation.

6.4 Future work15

This work has shown it is possible to develop a recharge model on a basinal-scale usinglargely remotely sensed data and to use distributed recharge modelling to interpretsubtle, seasonal variations in stored water mass indicated by GRACE data. Usinga recharge model it is possible to quantify how significant seasonal variations in thedifferent stored water masses within a large river basin are to the GRACE data. This20

level of interpretation is not possible using only bulk water balance models.Building from this initial application of the recharge model to interpret GRACE data,

future work is required to advance the calibration of the ZOODRM recharge model,to enable more accurate comparison of modelling results to the GRACE data. Futuredevelopment of ZOODRM recharge model should:25

4519





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

– Enable accurate simulation of run-off routing time of surface water within therecharge model across large distances (at present only annualised flows can becompared);

– Enable accurate simulation of stored water mass in surface water bodies,

– Include artificial storage of water in the man-made reservoirs in the recharge5

model, as well as known irrigation abstraction losses.

A further key need in future work is to calibrate the Nile recharge model to moderndischarge data. Obtaining modern discharge data for the Nile River and its tributariesis very difficult in the large transboundary basin where allocation of surface water re-sources is highly contentious. Techniques for using satellite radar altimetry to deter-10

mine surface water levels in lakes and rivers are being developed, and their applicationto the Nile Basin could provide modern estimates of river flows (e.g. Berry and Pinnock,2003) and improve the level of attainable model calibration. Whilst satellite altimetrytechniques would only generate estimated modern discharge data, the data would beindependent – a key factor in a large transboundary basin where water resources are15

under extreme pressure, and hydrological data highly contentious.

7 Conclusions

This study developed a groundwater recharge model for the Nile Basin, using theZOODRM code, and almost entirely remotely sensed input data. The recharge modelwas calibrated to observed hydrological data from the Nile, and then used to interpret20

the seasonal variation in terrestrial water storage indicated by GRACE data from theNile Basin. The results of this work are:

1. GRACE data for the Nile Basin indicates a seasonal spatial variation to gravityof ±0.005 mGal in the basin, which relates to annual change in terrestrial waterstorage change of approximately 200 km3: water input is rainfall, and much of this25

4520





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

water is evaporated within the basin since average annual outflow of the Nile isless than 30 km3.

2. The total annual groundwater recharge simulated by ZOODRM is 400 km3/yr– a figure considerably greater than river outflow from the Nile Basin. Withinthe semi-arid lower catchment average annual recharge is modelled to be 0–5

50 mm/yr, whilst in the sub-tropical upper catchment modelled recharge is on av-erage 250 mm/yr. These results are comparable to the few site studies of rechargein the basin.

3. Accounting for year-round discharge of groundwater, the seasonal change ingroundwater storage is modelled to be approximately 100–150 km3/yr within the10

Nile Basin, much greater than the seasonal change in soil moisture (30 km3/yr).Together, they account for between 50 and 90% of the annual water storage inthe catchment, the remainder being temporary water storage in surface waterand wetlands and irrigation.

4. The annual water mass variation (200 km3/yr) is an order of magnitude smaller15

than the rainfall input into the catchment (2000 km3/yr), and river outflow is neg-ligible (<30 km3/yr). This indicates the high evaporative losses to be significantwithin the Nile Basin. However, annual changes in atmospheric water volumesare inferred to be less than 10% of the total annual water change by the compar-ison of corrected and uncorrected GRACE data using the ECMWF model. This20

suggests that much of this evaporated water of is rapidly re-precipitated within thebasin, rather than the water being held in the atmosphere.

The results of this study show the value of using distributed recharge models, overbulk water balance models, to interpret subtle variations in GRACE data. Using agroundwater recharge model it is possible to gain an insight into the partitioning, and25

seasonal variation, of stored water masses within river basins. In addition, this workhas also shown it is possible to develop a plausible, robust, recharge model using

4521





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

almost entirely remotely sensed input data. Future work is required to advance thecalibration of the ZOODRM model, particularly in including a timing element in therun-off routing.

Acknowledgements. This paper is published with permission of the Director of the British Ge-ological Survey (Natural Environment Research Council).5

References

Abdalla, O. A. E.: Groundwater recharge/discharge in semi-arid regions interpreted from iso-tope and chloride concentrations in north White Nile Rift, Sudan, Hydrogeol. J., 17, 3, 679–692, 2009.

Anderson, O. B. and Hinderer, J.: Global inter-annual gravity changes from GRACE: Early10

results, Geophys. Res. Lett., 32, 1, L01402, doi:10.1029/2004GL020948, 2005.Bates, B., Kundzewicz, Z., Wu, S., and Palutikof, J. (Eds.): IPCC: Climate change and water,

IPCC Working Group II, Technical Paper of the Intergovernmental Panel on Climate Change,IPCC Secretariat, Geneva, 210 pp., 2007.

Bedada, T.: Combining space based GRACE gravity field measurement and climatologically av-15

eraged precipitation data to assess essential features of hydrological mass variations withinthe Nile Basin, MSc Thesis, Addid Ababa University, 2007.

Berry, P. A. and Pinnock, R. A.: The potential contribution of satellite altimetry to retrieval of theglobal hydrology runoff budget, Geophys. Res. Abstr., 5, abstract no. 04389, EGS-AGU-EUGJoint Assembly, Nice, France, April 2003.20

Beven, K. J.: Rainfall-runoff modelling: the primer, John Wiley and Sons, England, 2002.Brown, N. J. and Tregoning, P.: Quantifying GRACE data contamination effects on hydrological

analysis in the Murray-Darling Basin, southeast Australia, Aust. J. Earth. Sci., 57, 3, 329–335, 2010.

Bruinsma, S., Lemoine, J. M., Biancale, R., et al.: CNES/GRGS 10-day gravity field models25

(release 2) and their evaluation, Adv. Space Res., 45, 4, 587–601, 2010.Brunner, P. Hendricks Frannsen, H.-J., Kgotlhang, L., Bauer-Gottwein, P., and Kinzelbach, W.:

How can remote sensing contribute in groundwater modelling?, Hydrogeol. J., 15, 5–18,2007.

4522





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Conway, D.: From headwater tributaries to international river: Observing and adapting to cli-mate variability and change in the Nile Basin, Global Environ. Chang., 15, 99–114, 2005.

DFID: Water Storage and hydropower: supporting growth, resilience and low carbon devel-opment, DFID evidence-into-action paper, Department for International Development, UK,London, 26 pp., 2010.5

Doll, P., Kaspar, F., and Lehner, B.: A global hydrological model for deriving water availabilityindicators: model tuning and validation, J. Hydrol., 270, 105–134, 2003.

Doll, P. and Fiedler, K.: Global-scale modeling of groundwater recharge, Hydrol. Earth Syst.Sci., 12, 863–885, doi:10.5194/hess-12-863-2008, 2008.

FAO: Crop evapotranspiration; Guidelines for computing crop water requirements, Food and10

Agriculture Organization of the United Nations Irrigation and Drainage Paper 56, Rome,1998.

Farah, E. A., Mustafa, E. M. A., and Kumai. H.: Sources of groundwater recharge at confluenceof the Niles, Sudan, Environ. Geol., 39, 6, 667–672, 1999.

Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., et al.: The Shuttle Radar Topography Mission,15

Rev. Geophys., 45, RG2004, doi:10.1029/2005RG000183, 2007.Fontaine, B., Roucou, P., and Trzaska, S.: Atmospheric water cycle and moisture fluxes in the

West African Monsoon: mean annual cycles and relationship using NCEP/NCAR reanalysis,Geophys. Res. Lett., 30, 3, 1117–1121, 2002.

Fukuda, Y., Yamaoto, K., Hasegawa, T., Nakaegawa, T., Nishijima, J., and Taniguchi, M.: Mon-20

itoring groundwater variation by satellite and implications for in-situ gravity measurements,Sci. Tot. Environ., 407, 3173–3180, 2009.

Grey, D. and Sadoff, C. W.: Sink or Swim? Water security for growth and development, WaterPolicy, 9, 545–571, 2007.

Grindley, J.: The estimation of soil moisture deficits, Meteorol. Mag., 96, 1137, 97–108, 1967.25

Han, S. C., Shum, C. K., Bevis, M., et al.: Crustal dilation observed by GRACE after the 2004Sumatra-Andaman earthquake, Science, 313, 658–662, 2006.

Hughes, A. G., Mansour, M. M., and Robins, N.: Evaluation of distributed recharge in an uplandsemi-arid karst system: the West Bank Mountain Aquifer, Middle East, Hydrogeol. J., 16, 5,845–854, 2008.30

Jackson, C. R., Hughes, A. G., O Dochartaigh, B. E., Robins, N. S., and Peach D. W.: Numericaltesting of conceptual models of groundwater flow: A case study using the Dumfries BasinAquifer, Scot. J. Geol., 41(1), 51–60, 2005.

4523





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Karyabwite, D. R.: Water sharing in the Nile River Valley, UNEP/DEWA/GRID, Geneva, ProjectGNV011: Using remote sensing for the sustainable use of natural resources, 2000.

Lemonie, J.-M., Bruinsma, S., Loyer, S., Biancale, R., Marty, J.-C., Perosanz, F., and Balmino,G.: Temporal gravity field models inferred from GRACE data, Adv. Space Res., 39, 1620–1629, 2007.5

Lerner, D. N., Issar, A. S., and Simmers, I.: Groundwater recharge: a Guide to understandingand estimating natural recharge, IAH Publication, No. 8, 1990.

Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, J., Yang, L., and Merchant, J. W.:Development of a Global Land Cover Characteristics Database and IGBP DISCover from1-km AVHRR Data., Int. J. Remote. Sens., 21, 6–7, 1303–1330, 2000.10

MacDonald, A. M. and Calow, R. C.: Developing groundwater for secure water supplies inAfrica, Desalination, 248, 546–556, 2009.

MacDonald, A. M., Calow, R. C., Macdonald, D. M. J., Darling, W. G., O Dochartaigh, B. E.:What Impact will climate change have on rural water supplies in Africa?, Hydrolog. Sci. J.,54, 690–703, 2009.15

MacDonald, A. M., Darling, W. G., Nkotagu, H., Gooddy, D. C., Smedley, P. L., and Tyler-Whittle,R. A.: Environmental indicators of recharge to basement aquifers of Tanzania [abstract], in:Groundwater recharge assessment: are we any closer to an answer?, Geological Society,20–21 May 2009, University of East Anglia, 2009.

Mileham, L., Taylor, R., Thompson, J., Todd, M., and Tindimugaya, C.: Impact of rainfall distri-20

bution on the parameterisation of a soil-moisture balance model of groundwater recharge inequatorial Africa, J. Hydrol., 359, 46–58, 2008.

Mohammed, Y. A., Bastiaanssen, W. G. M., and Savenije, H. H. G.: Spatial variability of evap-oration and moisture storage in the swamps of the upper Nile studied by remote sensingtechniques, J. Hydrol., 289, 145–164, 2004.25

Mohamed, Y. A., van den Hurk, B. J. J. M., Savenije, H. H. G., and Bastiaanssen, W. G. M.:Hydroclimatology of the Nile: results from a regional climate model, Hydrol. Earth Syst. Sci.,9, 263–278, doi:10.5194/hess-9-263-2005, 2005.

Nicol, A.: The Nile: Moving beyond cooperation, UNESCO-IHP, Technical Documents in Hy-drology, PC-CP series, 16, 1–14, 2003.30

O Dochartaigh, B. E., MacDonald, A. M., Darling, W. G., Hughes, A. G., Li, J., and Shi, L.:Determining groundwater degradation from irrigation in desert-marginal northern China, Hy-drogeol. J., in press, 2010.

4524





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Penman, H. L.: Natural evaporation from open water, bare soil and grass, Proc. R. Soc. London,Series A, 193, 120–145, 1943.

Persits, F., Ahlbrandt, T., Tuttle, M., et al.: Maps Showing Geology, Oil and Gas Fields andGeological Provinces of Africa, USGS Open-file Report, 97-470A, 1997.

Rodell, M., Velicogna, I., and Famigletti, J. S.: Satellite-based estimates of groundwater deple-5

tion in India, Nature, 460, 999–1002, doi:10.1038/nature08238, 2009.Rodell, M., Famigletti, J. S., Chen, J. Seneviratne, S. I., Viterbo, P., Holl, S., and Wilson, C. R.:

Basin scale estimates of evapotranspiration using GRACE and other observations, Geophys.Res. Lett., 31, L20504, doi:10.1029/2004GL020873, 2004.

Schmidt, R, Schwintzer, P., Flechtner, F. Reigber, Ch., Guntner, A., Doll, P., Ramillien, G.,10

Cazenave, A., Petrovic, S., Jochmann, H., and Wunsch, J.: GRACE observations of changesin continental water storage, Global Planet. Change, 50, 112–126, 2006.

Sutcliffe, J. V. and Park, Y. P.: The Hydrology of the Nile, IAHS Special Publication, 5, IAHSPress, 1–160, 1999.

Sutcliffe, J. V.: Comment on ‘Spatial variability of evaporation and moisture storage in the15

swamps of the upper Nile studied by remote sensing techniques’ by Y. A. Mohamed et al.,2004, J. Hydrol., 314, 45–47, 2005.

Syed, T. H., Famiglietti, J. S., Chen, J., Rodell, M., Seneviratne, S. I., Viterbo, P., and Wilson,C. R.: Total basin discharge for the Amazon and Mississippi River basins from GRACE anda land-atmosphere water balance, Geophys. Res. Lett., 32, L24404, 2005.20

Tapley, B. D., Bettadpur, S., Ries, J. C., et al.: GRACE measurements of mass variability in theEarth system, Science, 305, 503–505, 2004.

Taylor, R. and Howard, K.: Groundwater recharge in the Victoria Nile basin of East Africa:support for the soil-moisture balance method using stable isotope and flow modelling studies,J. Hydrol., 180, 31–53, 1990.25

UNESCO: Africa Geological Map Scale (1:5 000 000), 6 Sheets, UNESCO, Paris, 1991.Vallet-Coulomb, C., Legesse, D., Gasse, F., Travi, Y., and Chernet, T.: Lake evaporation esti-

mates in tropical Africa (Lake Ziway, Ethiopia), J. Hydrol., 245, 1–18, 2001.

4525





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Table 1. Calibration results of the ZOODRM model based on modelled annualised dischargein 2003.

Annual discharge(km3/yr)

Aswan Dongola Bahr el Jebel Lake Victoria

Observed 58.4 69.0 48.3 38.5Modelled 60.1 80.6 69.1 42.6

4526





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 1. Major sub-catchments of the Nile Basin: Victoria Nile, Sudd wetland, White Nile, So-bat, Blue Nile, Atbara and the Main Nile. Annual hydrographs indicate the seasonal pulse ofdischarge in the Main Nile river, downstream of the Sobat, Blue Nile and Atbara tributaries.In contrast outflow from the Sudd provides a constant annual baseflow to the Nile. Note allhydrographs display a typical hydrological year, and all discharge scales on the plots are thesame.

4527





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 2. GRACE data for the Nile Basin, inferred to relate to changes in terrestrial water storagemass changes. Each month of GRACE data reflects the change in total stored water mass,relative to stored water mass in March 2004 (the driest hydrological month of the modelledperiod).

4528





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 3. Comparison of monthly rainfall, to the monthly river outflow to Mediterranean Sea, andmonthly GRACE data (presented as water volume relative to March 2004) for 2004.

4529





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 4. Monthly volumetric estimates of the partitioning of rainfall within the Nile Basin betweenthe main components of the water cycle. Soil moisture is not displayed on the plot, because theZOODRM model only outputs the monthly change in soil moisture water mass, rather than totalmonthly soil water mass – a direct comparison to the other modelling results would thereforebe invalid.

4530





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 5. Average annual groundwater recharge map of the Nile Basin. The data points andvalues on the recharge map represent annual recharge values determined from local studiesby other authors based on isotope and chloride concentrations in the groundwater.

4531





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 6. Comparison of the Nile GRACE data with the modelled annual variation in groundwaterstorage and soil moisture. The groundwater and soil moisture storage changes are cumulatedvalues which have been normalised to March 2004, like the GRACE data.

4532





HESSD7, 4501–4533, 2010



H. C. Bonsor et al.

Title Page



Tables Figures

J I

J I

Back Close

Full Screen / Esc



Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|D

iscussionP

aper|

Fig. 7. Comparison of GRACE data for the Nile Basin to the total water volume stored in soilmoisture and groundwater recharge. In drier years (e.g. 2004) when the volume of seasonalrunoff water and surface water storage is less, the total water storage mass variation in thebasin, is comprised predominantly of seasonal change in soil moisture and groundwater stor-age.

4533



